context-optimizer
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses the
@xenova/transformerslibrary, which downloads the 'all-MiniLM-L6-v2' embedding model from Hugging Face during initialization to support semantic pruning and archival indexing. - [PROMPT_INJECTION]: The skill ingests, archives, and re-injects conversation history into the agent's active context. Because this history includes untrusted user input, it represents an indirect prompt injection surface where instructions embedded in previous messages could be persisted in the local archive and later re-introduced into the agent's prompt.
- [COMMAND_EXECUTION]: Provides a CLI tool (scripts/cli.js) and installation commands (npm install) intended for standard setup and local message processing, with no evidence of unauthorized or hidden command execution.
Audit Metadata